| { | |
| "model_type": "ultravox", | |
| "architectures": [ | |
| "UltravoxModel" | |
| ], | |
| "audio_token_id": 151648, | |
| "audio_config": { | |
| "model_type": "vibevoice_audio_vae", | |
| "hidden_size": 1536, | |
| "sampling_rate": 24000, | |
| "compress_ratio": 3200 | |
| }, | |
| "text_config": { | |
| "model_type": "qwen2", | |
| "hidden_size": 1536, | |
| "intermediate_size": 8960, | |
| "num_hidden_layers": 28, | |
| "num_attention_heads": 12, | |
| "num_key_value_heads": 2, | |
| "vocab_size": 151936, | |
| "max_position_embeddings": 65536, | |
| "rope_theta": 1000000.0, | |
| "rms_norm_eps": 1e-06, | |
| "tie_word_embeddings": true, | |
| "bos_token_id": 151643, | |
| "eos_token_id": 151645 | |
| }, | |
| "transformers.js_config": { | |
| "kv_cache_dtype": { | |
| "q4f16": "float16", | |
| "fp16": "float16" | |
| }, | |
| "dtype": { | |
| "audio_encoder": "q4", | |
| "embed_tokens": "q4", | |
| "decoder_model_merged": "q4" | |
| } | |
| } | |
| } |